CN112884190A

CN112884190A - Flow prediction method and device

Info

Publication number: CN112884190A
Application number: CN201911205351.5A
Authority: CN
Inventors: 钟晓超
Original assignee: Hangzhou Hikvision Digital Technology Co Ltd
Current assignee: Hangzhou Hikvision Digital Technology Co Ltd
Priority date: 2019-11-29
Filing date: 2019-11-29
Publication date: 2021-06-01
Anticipated expiration: 2039-11-29
Also published as: CN112884190B

Abstract

The embodiment of the invention provides a flow prediction method and a flow prediction device. Wherein the method may comprise: acquiring characteristics of point positions to be predicted on multiple dimensions, wherein the characteristics on the multiple dimensions at least comprise topological graph structure characteristics and historical flow characteristics; inputting the characteristics of the point location to be predicted on the multiple dimensions to a flow prediction model to obtain a flow prediction result of the point location to be predicted, wherein the flow prediction model is trained in advance by sample data labeled with truth values, and the sample data are the characteristics of the point location to be predicted and the other point locations labeled with truth values on the multiple dimensions. The flow prediction model can learn the mapping from the characteristics of different point positions to the flow integrally through the structural characteristics of the topological graph instead of learning the mapping from the characteristics of different point positions to the flow discretely. The learned traffic to feature mapping is more accurate because more sample data is referenced.

Description

Flow prediction method and device

Technical Field

The invention relates to the technical field of intelligent traffic, in particular to a flow prediction method and a flow prediction device.

Background

In the related technology, the traffic at the future time can be predicted based on historical traffic data, so that the traffic planning can be performed in a targeted manner. However, since different points are spatially discrete, in the related art, when predicting the traffic of one point, the traffic is often based on only the historical traffic data of the point.

However, historical traffic data of a point is often limited, that is, data referred to when predicting traffic may not be sufficient, and thus predicted traffic may not be accurate enough.

Disclosure of Invention

The embodiment of the invention aims to provide a flow prediction method and a flow prediction device so as to improve the accuracy of predicted flow. The specific technical scheme is as follows:

in a first aspect of the embodiments of the present invention, a traffic prediction method is provided, where the method includes:

acquiring characteristics of a point to be predicted on multiple dimensions, wherein the characteristics on the multiple dimensions at least comprise topological graph structure characteristics and historical flow characteristics, the topological graph structure characteristics are used for representing the communication relation between the point to be predicted and other points in a topological graph structure, and the topological graph structure is constructed on the basis of internet of things data acquired at the point to be predicted and the other points and is used for representing the relevance of the point to be predicted and the other points on a time dimension and/or a space dimension;

inputting the characteristics of the point location to be predicted on the multiple dimensions to a flow prediction model to obtain a flow prediction result of the point location to be predicted, wherein the flow prediction model is trained in advance by sample data labeled with truth values, and the sample data are the characteristics of the point location to be predicted and the other point locations labeled with truth values on the multiple dimensions.

In a possible embodiment, the topological graph structure is constructed by:

acquiring a plurality of pieces of internet of things data acquired from the point location to be predicted and the other point locations, wherein each piece of internet of things data comprises data points in a plurality of dimensions, and at least comprises data points in a time dimension and data points in a space dimension;

constructing a plurality of nodes according to the fractional data of the plurality of internet of things data on the spatial dimension, wherein each node is used for representing the point location to be predicted and one point location of the other point locations;

determining the correlation between every two nodes in the plurality of nodes, wherein the correlation is used for representing the correlation degree between the data of the internet of things corresponding to one node in the two nodes and the data of the internet of things corresponding to the other node in the spatial dimension and the time dimension;

and according to the correlation among the nodes, constructing edges among the nodes to obtain a topological graph structure.

In one possible embodiment, the determining the correlation between each two nodes of the plurality of nodes comprises:

for each two nodes in the plurality of nodes, determining the interval of the data of the internet of things corresponding to one node in the two nodes and the data of the internet of things corresponding to the other node in the space dimension and/or the time dimension;

and if the interval is smaller than a preset interval threshold value, determining that the correlation between the two nodes is related.

In a possible embodiment, the flow prediction model is trained in advance by:

for each node in the topological graph structure, determining the topological graph structure characteristics of the node;

for each piece of internet-of-things data in the plurality of pieces of internet-of-things data, merging the topological graph structure characteristics of the node corresponding to the piece of internet-of-things data with the piece data of the piece of internet-of-things data in other dimensions except the space dimension and the flow dimension, and taking the piece data of the piece of internet-of-things data in the flow dimension as a marking value to obtain sample data of the node;

for each node in the topological graph structure, sequencing sample data of the node according to the time sequence to obtain a sample sequence;

performing sliding window processing on the sample sequence to obtain a plurality of time sequence data;

and training to obtain a flow prediction model by utilizing the plurality of time sequence data of each node in the topological graph structure.

In a possible embodiment, the training to obtain a traffic prediction model by using the time series data of each node in the topological graph structure includes:

respectively training a plurality of basic models by utilizing the fractional data of the plurality of time sequence data of each node in the topological graph structure on different dimensions;

and fusing the trained basic models to obtain a flow prediction model.

In a second aspect of the embodiments of the present invention, there is provided a trajectory prediction method, including:

determining a plurality of historical point positions passed by a target object as a sequence to be predicted;

obtaining features of the plurality of historical point locations and the plurality of candidate point locations, wherein the features at least comprise topological graph structure features, the topological graph structure features are used for representing the connection relationship between the point location and other point locations in a topological graph structure, and the topological graph structure is constructed based on historical tracks of the target object and at least one other object and is used for representing the relevance of the plurality of historical point locations and the plurality of candidate point locations in a time dimension and/or a space dimension;

and taking the candidate point location with the matched features of the plurality of candidate point locations and the features of the plurality of historical point locations as the point location at the next moment of the sequence to be predicted.

In a possible embodiment, the taking, as a point of the next time point of the sequence to be predicted, a candidate point of the plurality of candidate points whose feature matches the feature of the plurality of historical points, includes:

for each candidate point location in the plurality of candidate point locations, calculating a similarity of the features of the candidate point location and the features of the plurality of historical point locations;

and taking the candidate point position with the highest similarity in the plurality of candidate point positions as the next moment point position of the sequence to be predicted.

In a possible embodiment, the topological graph structure is constructed by:

acquiring a plurality of bayonet data of the target object and at least one other object;

aggregating point locations corresponding to two pieces of bayonet data belonging to the same time window and the same object in the plurality of bayonet data to obtain historical tracks of the target object and at least one other object;

constructing a topological graph structure based on point location pairs formed by adjacent point locations in the historical track of the target object to obtain an individual topological graph structure;

constructing a topological graph structure based on a point location pair consisting of adjacent point locations in the historical track of the at least one other object to obtain a group topological graph structure;

and fusing the individual topological graph structure and the group topological graph structure.

In a third aspect of the embodiments of the present invention, there is provided a fake-licensed vehicle detection method, including:

acquiring a historical track of a vehicle to be detected;

constructing a topological graph structure based on historical track point positions in the historical tracks;

calculating the number of connected subgraphs in the topological graph structure, wherein the connected subgraphs are subsets of nodes and edges in the topological graph structure, and any two nodes in each connected subgraph are connected through the edges in the connected subgraph;

determining whether the number is 1;

and if the number is not 1, determining that the vehicle to be detected is a fake plate vehicle.

In a possible embodiment, the method further comprises:

and if the number is 1, determining that the vehicle to be detected is not a fake-licensed vehicle.

In a possible embodiment, the constructing a topological graph structure based on historical track point locations in the historical tracks includes:

determining the frequency of the historical track point in the historical track aiming at each historical track point in the historical track;

determining whether the times of the historical track point locations are greater than a preset time threshold;

if the times are larger than a preset time threshold value, taking the historical track point location as a hot historical track point location;

and constructing a topological graph structure based on all the hot historical track point positions.

In a fourth aspect of the embodiments of the present invention, there is provided a flow rate prediction apparatus, including:

the characteristic obtaining module is used for obtaining characteristics of the point to be predicted on multiple dimensions, wherein the characteristics on the multiple dimensions at least comprise topological graph structure characteristics and historical flow characteristics, the topological graph structure characteristics are used for representing the communication relation between the point to be predicted and other point positions in a topological graph structure, and the topological graph structure is constructed on the basis of the internet of things data collected at the point to be predicted and the other point positions and is used for representing the relevance of the point to be predicted and the other point positions on the time dimension and/or the space dimension;

and the flow prediction module is used for inputting the characteristics of the point location to be predicted on the multiple dimensions into a flow prediction model to obtain a flow prediction result of the point location to be predicted, the flow prediction model is trained by sample data marked with truth values in advance, and the sample data is the characteristics of multiple point locations marked with truth values on the multiple dimensions in the point location to be predicted and other point locations.

In a possible embodiment, the topological graph structure is constructed by:

In a possible embodiment, the flow prediction model is trained in advance by:

and fusing the trained basic models to obtain a flow prediction model.

In a fifth aspect of embodiments of the present invention, there is provided a trajectory prediction apparatus, including:

the prediction sequence module is used for determining a plurality of historical point positions passed by the target object as a sequence to be predicted;

the characteristic expression module is used for acquiring characteristics of the historical point positions and the candidate point positions, the characteristics at least comprise topological graph structural characteristics, the topological graph structural characteristics are used for representing the connection relation between the point position and other point positions in a topological graph structure, and the topological graph structure is constructed on the basis of historical tracks of the target object and at least one other object and is used for representing the relevance of the historical point positions and the candidate point positions in a time dimension and/or a space dimension;

and the feature matching module is configured to use, as a next time point of the sequence to be predicted, a candidate point of the plurality of candidate points, where the feature is matched with the feature of the plurality of historical points.

In a possible embodiment, the feature matching module is specifically configured to, for each candidate point location in the plurality of candidate point locations, calculate a similarity between the feature of the candidate point location and the features of the plurality of historical point locations;

In a possible embodiment, the topological graph structure is constructed by:

In a sixth aspect of an embodiment of the present invention, there is provided a fake-licensed vehicle detection apparatus, including:

the track acquisition module is used for acquiring the historical track of the vehicle to be detected;

the topological graph construction module is used for constructing a topological graph structure based on historical track point positions in the historical tracks;

the connected subgraph calculation module is used for calculating the number of connected subgraphs in the topological graph structure, the connected subgraphs are subsets of nodes and edges in the topological graph structure, and any two nodes in each connected subgraph are connected through the edges in the connected subgraph;

the judging module is used for determining whether the number is 1;

and the detection module is used for determining that the vehicle to be detected is a fake plate vehicle if the number is not 1.

In a possible embodiment, the detection module is further configured to determine that the vehicle to be detected is not a fake-licensed vehicle if the number is 1.

In a possible embodiment, the topological graph constructing module is specifically configured to, for each historical track point location in the historical tracks, determine the number of times that the historical track point location appears in the historical tracks; determining whether the times of the historical track point locations are greater than a preset time threshold; if the times are larger than a preset time threshold value, taking the historical track point location as a hot historical track point location; and constructing a topological graph structure based on all the hot historical track point positions.

In a seventh aspect of the embodiments of the present invention, there is provided an electronic apparatus, including:

a memory for storing a computer program;

a processor adapted to perform the method steps of any of the above first aspects when executing a program stored in the memory.

In an eighth aspect of embodiments of the present invention, there is provided an electronic apparatus, including:

a memory for storing a computer program;

a processor for implementing the method steps of the second aspect when executing the program stored in the memory.

In a ninth aspect of embodiments of the present invention, there is provided an electronic apparatus, including:

a memory for storing a computer program;

a processor for implementing the method steps of any of the above third aspects when executing a program stored in the memory.

In a tenth aspect of embodiments of the present invention, there is provided a computer-readable storage medium having stored therein a computer program which, when executed by a processor, performs the method steps of any one of the above-mentioned first aspects.

In an eleventh aspect of embodiments of the present invention, there is provided a computer-readable storage medium having stored therein a computer program which, when executed by a processor, performs the method steps of any one of the above-mentioned second aspects.

In a twelfth aspect of embodiments of the present invention, there is provided a computer-readable storage medium having stored therein a computer program which, when executed by a processor, implements the method steps of any one of the above-mentioned third aspects.

The traffic prediction method and the traffic prediction device provided by the embodiment of the invention can enable the traffic prediction model to learn the mapping from the features of different point locations to the traffic in a whole manner through the structural features of the topological graph instead of learning the mapping from the features of different point locations to the traffic discretely. The learned traffic to feature mapping is more accurate because more sample data is referenced. . Of course, not all of the advantages described above need to be achieved at the same time in the practice of any one product or method of the invention.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

Fig. 1 is a schematic flow chart of a traffic prediction method according to an embodiment of the present invention;

fig. 2 is a schematic flow chart of a method for constructing a topology structure according to an embodiment of the present invention;

fig. 3 is a schematic flow chart of a method for constructing a traffic prediction model according to an embodiment of the present invention;

fig. 4 is a schematic flow chart of a traffic prediction model training method according to an embodiment of the present invention;

FIG. 5 is a flowchart illustrating a trajectory prediction method according to an embodiment of the present invention;

fig. 6 is another schematic flow chart of a method for constructing a topology structure according to an embodiment of the present invention;

FIG. 7 is a schematic flow chart of a method for detecting a fake-licensed vehicle according to an embodiment of the present invention;

FIG. 8a is a schematic diagram of a topology provided by an embodiment of the present invention;

FIG. 8b is another schematic diagram of a topology provided by an embodiment of the present invention;

fig. 9 is another schematic flow chart of a method for constructing a topology structure according to an embodiment of the present invention;

fig. 10 is a schematic structural diagram of a flow prediction apparatus according to an embodiment of the present invention;

FIG. 11 is a schematic structural diagram of a trajectory prediction apparatus according to an embodiment of the present invention;

fig. 12 is a schematic structural diagram of a fake-licensed vehicle detecting device according to an embodiment of the present invention;

fig. 13 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Referring to fig. 1, fig. 1 is a schematic flow chart of a flow prediction method according to an embodiment of the present invention, which may include:

s101, obtaining characteristics of the point to be predicted on multiple dimensions.

The characteristics on the multiple dimensions at least comprise topological graph structure characteristics and historical flow characteristics. The topological graph structure characteristic is used for representing the connection relation between the point to be predicted and other point positions in the topological graph structure. The topological graph structure is constructed on the basis of the data of the internet of things collected by the point location to be predicted and other point locations and is used for representing the relevance of the point location to be predicted and other point locations on the time dimension and/or the space dimension.

The topological graph structure comprises a plurality of nodes, wherein each node is used for representing a point position to be predicted and one of other point positions. The construction of the topology structure will be described in detail in the following embodiments, and will not be described herein. It can be understood that, because the topological graph structure features can reflect the association relationship between the point location to be predicted and other point locations in the time dimension and/or the space dimension, and the association relationship between one point location and other point locations in the time dimension and/or the space dimension can be regarded as one piece of information of the point location, and the information is determined based on the internet of things data of the point location to be predicted and the internet of things data of other point locations, the topological graph structure features of the point location to be predicted can reflect the features of the internet of things data acquired from other point locations to a certain extent.

The features in multiple dimensions may include other features according to different actual requirements, besides the topological graph structure feature and the historical flow rate feature, which is not limited in this embodiment. For example, in one possible embodiment, seven-dimensional features may be included, the first feature being a topological structure feature, the second feature being a historical traffic feature, the third feature representing time information, the fourth feature representing geographical location information, the fifth feature representing weather information, the sixth feature representing event information for events occurring within a historical time period, and the seventh feature representing other information.

And S102, inputting the characteristics of the point location to be predicted on multiple dimensions into a flow prediction model to obtain a flow prediction result of the point location to be predicted.

The flow prediction model is trained by sample data labeled with truth values in advance, and the sample data is the characteristic of a plurality of dimensions labeled with truth values of the point positions to be predicted and a plurality of point positions in other point positions. Since the sample data is labeled with a true value, the sample data has one more dimension compared with the features in the multiple dimensions.

For example, the sample data of the ith point may be X_i＝{x_id1，x_id2，x_id3，x_id4，x_id5，x_id6，x_id7|y_iIn the form of (a) }, wherein, y_iIs the true value of the label, i.e. the flow of the point at a certain moment, x_id1As the first feature of the ith point at that time, x_id2And (e) repeating the above steps for the second feature of the ith point at the time point (for the features referred to by the first feature and the second feature, reference may be made to the relevant description in S101, and details are not described here again).

The flow prediction model is used to implement a mapping of features in multiple dimensions to flow. The flow prediction model may be a neural network model obtained based on deep learning, or may be a model obtained based on traditional machine learning, which is not limited in this embodiment.

As analyzed in the foregoing, the topological graph structure features can reflect, to some extent, features of the data of the internet of things collected from other point locations. Therefore, the flow prediction model is trained by using the characteristics of a plurality of dimensions including the topological graph structure characteristics, so that the flow prediction model can learn the mapping from the characteristics of different point positions to the flow integrally through the topological graph structure characteristics instead of learning the mapping from the characteristics of different point positions to the flow discretely. The learned traffic to feature mapping is more accurate because more sample data is referenced.

The following describes the construction of a topological graph structure, and referring to fig. 2, fig. 2 is a schematic flowchart of a method for constructing a topological graph structure according to an embodiment of the present invention, and the method may include:

s201, acquiring a plurality of Internet of things data collected at the point position to be predicted and other point positions.

Wherein each internet of things data comprises score data in multiple dimensions and at least comprises score data in a time dimension and score data in a space dimension. The data of the internet of things can be data collected by entity equipment, and the entity equipment can be different according to different application scenes, for example, the data of the internet of things can comprise speed data obtained by radar measurement, image data collected by a camera, electronic identification information read by an electronic identification reader-writer, and one or more of GPS positioning information collected by a GPS positioning device. The data collected by the entity devices may be multi-dimensional, for example, the dimensions of the data collected by the cameras may include a collection location (spatial dimension), a collection time (temporal dimension), personnel and/or vehicle image information.

For convenience of description, the following will take the prediction of traffic flow as an example, and the same principle is applied to the application scenario of people flow prediction, which is not described again. One piece of internet-of-things data may be { point a, 9:30, zhe AXXXXXXX, red, vendor 1, model 1}, which represents a red vehicle of model 1 produced by vendor 1 with a license plate number of zhe AXXXXXXX, and appears at point a at 9:30, where point a is point data of the internet-of-things data in a spatial dimension, and 9:30 is point data of the internet-of-things data in a time dimension.

In a possible embodiment, vehicle information of passing vehicles is collected at a point location to be predicted and other point locations respectively to obtain passing data of each point location, each passing data at least comprises time (i.e. point data in a time dimension) and location (i.e. point data in a space dimension) of collecting the vehicle information, the collected passing data is cleaned, the passing data lacking necessary information (such as license plate numbers), redundant and abnormal passing data is removed, and the remaining passing data is used as a plurality of internet of things data.

S202, constructing a plurality of nodes according to the fractional data of the internet of things data in the space dimension.

Each node is used for representing a point location to be predicted and one of other point locations, and the point locations represented by different nodes are different.

S203, determining the correlation between every two nodes in the plurality of nodes.

The correlation is used for representing the correlation degree between the data of the internet of things corresponding to one node of the two nodes and the data of the internet of things corresponding to the other node in the space dimension and the time dimension. The expression manner of the correlation may be different according to different application scenarios, for example, the correlation may be associated or not associated, or may be a numerical value with a value range of [0,1] for representing the degree of association between two nodes, for example, 1 represents that the two nodes are completely associated, 0 represents that the two nodes are completely unrelated, and the numerical value of (0,1) represents that the two nodes are partially associated, and the higher the numerical value is, the higher the degree of association between the two nodes is.

For example, the internet of things data corresponding to one node is determined, the interval of the internet of things data corresponding to another node in the space dimension and/or the time dimension is determined, and if the interval is smaller than a preset interval threshold, the correlation between the two nodes is determined to be associated.

For example, assume that one node is used to represent point a, and point a corresponds to internet of things data { point a, 9:30, zhe AXXXXXXX, red, vendor 1, model 1}, and another node is used to represent point B, and point B corresponds to internet of things data { point B, 10:00, zhe AXXXXXXX, red, vendor 1, model 1 }. It may be calculated that the interval between the two pieces of internet-of-things data in the time dimension is 30 minutes, and assuming that the preset interval threshold is 1 hour, it may be determined that the two nodes are associated. Or, the spatial distance between the point location a and the point location B may be calculated, and if the preset interval threshold is 1 km, the two nodes may be considered to be unrelated.

In other possible embodiments, it may also be determined whether two nodes are related according to actual requirements, or by integrating the intervals in the time dimension and the space dimension. In other possible embodiments, a value with a value range of [0,1] may also be obtained by calculating according to a preset algorithm with the interval in the time dimension and/or the space dimension, as the correlation between the two nodes, for example, the inverse of the interval in the time dimension or the space dimension may be calculated as the correlation between the two nodes. The present embodiment does not limit this.

And S204, constructing edges among the nodes according to the correlation among the nodes to obtain a topological graph structure.

The method of constructing the edge may be different according to different application scenarios, and for example, the method may be to construct an edge with an edge weight, or may be to construct an edge without an edge weight, where the constructed edge may be a directional edge with a direction, or may be a non-directional edge without a direction.

For example, an edge may be used to connect associated nodes in the plurality of nodes, and for an application scenario in which a directed edge is constructed, the direction of the edge may be determined according to the sequence of the internet of things data corresponding to the two connected nodes in the time dimension. For example, assume that a node a is associated with a node B, the node a corresponds to internet of things data { point a, 9:30, zhe AXXXXXXX, red, vendor 1, model 1}, the node B corresponds to internet of things data { point B, 10:00, zhe AXXXXXXX, red, vendor 1, model 1}, and since the point data of the internet of things data corresponding to the node a is 9:30 and the point data of the internet of things data corresponding to the node B is 10:00 in the time dimension, the internet of things data corresponding to the node a precedes the internet of things data corresponding to the node B, and thus an edge pointing to the node B from the node a may be constructed.

It can be understood that, in the constructed topological graph structure, an edge may reflect a time and/or space association relationship between one node and other nodes, and a time and/or space association relationship between one node and other nodes may be regarded as information of the node, and the information is determined based on the data of the internet of things corresponding to the node and the data of the internet of things corresponding to other nodes except the node. Therefore, in the topological graph structure, the data utilized by each node is more comprehensive, and the real information of the node can be more accurately reflected.

Referring to fig. 3, fig. 3 is a schematic flow chart of a method for constructing a flow prediction model according to an embodiment of the present invention, where the method may include:

s301, determining the topological graph structure characteristics of each node in the topological graph structure.

The topological graph structure feature is used for representing the connection relationship between the node and other nodes in the topological graph structure. It is understood that two nodes may be considered similar in their positions in the topology if the topology features of the two nodes are similar. Because each node represents one point location, the topological graph structure feature can also be used for representing the connection relationship of two point locations in the topological graph structure.

And S302, aiming at each piece of Internet of things data in the plurality of pieces of Internet of things data, merging the topological graph mechanism characteristics of the topological graph nodes of the nodes corresponding to the piece of Internet of things data with the piece data of the Internet of things data in other dimensions except the space dimension and the flow dimension, and taking the piece data of the Internet of things data in the flow dimension as a marking value to obtain sample data of the node.

Because the dimensions included in the internet of things data are different in different application scenarios and are not described conveniently, it is assumed that the ith internet of things data can be represented as D_i＝{x_i1，x_i2，x_i3，x_i4，x_i5，x_i6，x_i7，y_iIn which x_i1Represents the monitoring point position, x, of the collected data of the Internet of things_i2Representing the flow characteristics, x, of the monitoring point location in the historical time period_i3Time information, x, representing the current time of the point being monitored_i4Geographical location information, x, representing the location of the monitoring point_i5Weather information, x, representing the current time of the point location_i6Event information x representing an event occurring within the point site history period_i7Other information indicating the current time of the point being monitored, y_iAnd the traffic flow of the current time of the monitoring point is shown.

The sample data of the ith node may be represented as X_i＝{x_id1，x_id2，x_id3，x_id4，x_id5，x_id6，x_id7|y_iIn which X_iSample data for the ith node, x_id1Is a topological graph structure characteristic of the node, x_id2Data of the internet of things for the node is in x_i2Characteristic of (1), x_id3Data of the internet of things for the node is in x_i3The above feature, analogized in turn, y_iMay be the annotation data of the sample data of the node.

And S303, sequencing the sample data of each node in the topological graph structure according to a time-first-hand sequence to obtain a sample sequence.

For convenience of description, assume that the ith node is at t_jSample data of time X_ijAnd assume t₁＜t₂…＜t_nThen the sample data for that node may be arranged into a sample sequence { X }_i1，X_i2，…X_in}。

S304, performing sliding window processing on the sample sequence to obtain a plurality of time sequence data.

The width of the window selected in the sliding window processing may be different according to the application scenario. For example, assuming that the window is four sample sequences wide, the sample data included in the initial position of the window is X_i1、X_i2、X_i3And X_i4After the window slides by one unit, the sample data included in the window is X_i2、X_i3、X_i4And X_i5And so on until the window slides to the end of the sequence. Taking the sample data included after each window sliding and the sample data initially included as a time sequence data, and obtaining the time sequence data as follows:

{[X_i1,X_i2,X_i3.X_i4],[X_i2,X_i3,X_i4.X_i5],[X_in-3,X_in-2,X_in-1.X_in]}

each time series data can be regarded as a sub-sample data of the node sample data before the sliding window processing, so that the time series data can also be used for constructing the model, and the principle of constructing the prediction model according to the node sample data is the same as that of constructing the prediction model according to the node sample data. By adopting the embodiment, more data for constructing the prediction model can be obtained through sliding window processing, so that the obtained prediction model is more accurate.

Taking the predicted traffic flow as an example, assuming that sample data of a node may include characteristics of traffic data acquired within 10 days from a point represented by the node, if a prediction model is directly constructed based on the sample data, the constructed prediction model is a model for predicting the traffic flow based on the traffic data within 10 days. And each time series data obtained through sliding window processing may include the characteristics of traffic data acquired within 4 days of a point represented by a node, and if a prediction model is constructed based on the time series data, the constructed prediction model is a model for predicting the traffic flow based on the traffic data within 4 days.

S305, training to obtain a flow prediction model by using a plurality of time sequence data of each node in the topological graph structure.

F () represents the flow prediction model to be constructed, and a time sequence data is assumed to be x_id1，x_id2，x_id3，x_id4，x_id5，x_id6，x_id7|y_i}, the following relationship exists:

f(x_id1，x_id2，x_id3，x_id4，x_id5，x_id6，x_id7)＝y_i

according to the time sequence data of the nodes and the relation, f () can be obtained through training in a machine learning and/or deep learning mode.

As described in relation to S305, the process of training the traffic prediction model may be considered to be based on f (x)_id1，x_id2，x_id3，x_id4，x_id5，x_id6，x_id7)＝y_iFitting is performed on f (). In an alternative embodiment, a mathematical model of f () may be selected, and the parameters of the mathematical model may be determined by fitting. However, in some application scenarios, because the difference of the data of the internet of things in different dimensions may be large, it is difficult to determine a suitable mathematical model, and then a prediction model is obtained through fitting.

In view of this, referring to fig. 4, fig. 4 is a schematic flow chart of a flow prediction model training method according to an embodiment of the present invention, which may include:

s401, respectively training a plurality of basic models according to data with different dimensions in a plurality of time sequence data of each node in the topological graph structure.

The basic model may include KNN (K-Nearest Neighbor classification), randomtrees, LightGBM (learning algorithm based on decision tree), etc., and in an alternative embodiment, CNN (Convolutional Neural Network) and LSTM (Long Short-Term Memory Network) may be used as the basic model.

Still with X_i＝{x_id1，x_id2，x_id3，x_id4，x_id5，x_id6，x_id7|y_iFor example, assuming that the base model of CNN is u (), and the base model of LSTM is v (), the respective training may be to adjust the parameters in u () and v () respectively based on the following relations:

u(x_id1，x_id2，x_id3，x_id4)＝y，v(x_id5，x_id6，x_id7)＝y。

s402, fusing the trained basic models to obtain a flow prediction model.

The fusion mode may be different according to different application scenarios. Illustratively, still taking the example in S301 as an example, assuming that the prediction model obtained after the fusion is f (), f () can be determined in any one of the following manners:

the first method is as follows: f () + α u (), + β v ();

the second method comprises the following steps: f (), u (v ()).

Alpha and beta are preset weights.

By adopting the embodiment, the constructed prediction model can be well fitted with the sample data at a plurality of different latitudes by fusing a plurality of different technical models, namely, the constructed prediction model is more accurate.

It is understood that, in a possible embodiment, the construction of the topological graph mechanism and the training of the flow prediction model may be completed by the execution subject of the flow prediction method, or may be completed by other electronic devices with computing capability besides the execution subject, which is not limited in this embodiment.

In some application scenarios, in order to effectively manage the target object, the trajectory of the target object needs to be predicted, and in a possible implementation manner, the possible degree that each point is the point at the next time of the target object may be calculated according to a rule learned from the historical trajectory of the target object and a point pair included in the trajectory to be predicted of the target object, and the point with the highest possible degree is used as the point at the next time of the target object. But the point pairs tend to reflect only the correlation of the points in the trace in the time domain,

in view of this, an embodiment of the present invention provides a trajectory prediction method, referring to fig. 5, where fig. 5 is a schematic flow chart of the trajectory prediction method provided in the embodiment of the present invention, and the method may include:

s501, determining a plurality of historical point positions passed by the target object as a sequence to be predicted.

The target object may be a vehicle, a human or other kind of object according to different application scenarios, which is not limited in this embodiment. For convenience of description, the following description will take vehicle trajectory prediction as an example, and the principles of trajectory prediction for other types of objects are the same, and therefore will not be described again.

The method may be to obtain a plurality of history point locations that a target object passes through within a preset history time window, and may be, for example, to read gateway data acquired by each gateway within 12:00-14:00 of month 1, where the gateway data may be represented in a key-id-value form, where the key is a vehicle identifier of a vehicle corresponding to the gateway data, id is a gateway identifier of a gateway corresponding to the acquired gateway data, and value is time for acquiring the gateway data. For example, vehicle 1-gateway 1-12:10 may indicate that 12:10 minutes of vehicle 1 are present at gateway 1.

The gateway data in the gateway data group are sorted in the order of values from small to large, the value of the gateway data in the gateway data group may be larger when the represented time is later, or the value of the vlan when the represented time is earlier may be larger, and this is not limited in this embodiment. For two adjacent gateway data in the gateway data group, if the id of the two gateway data is the same and the difference value of value is smaller than a preset difference threshold, combining the two gateway data until there are no two adjacent gateway data with the same id and the difference value of value smaller than the preset difference threshold. In this case, the gateway data set can be regarded as a sequence to be predicted.

S502, obtaining characteristics of a plurality of historical point positions and a plurality of candidate point positions.

The characteristic of each point location at least comprises a topological graph structural characteristic of the point location, the topological graph structural characteristic is used for representing the connection relation between the point location and other point locations in a topological graph structure, and the topological graph structure is constructed based on the historical tracks of the target object and at least one other object and is used for representing the relevance of a plurality of historical point locations and a plurality of candidate point locations in a time dimension and/or a space dimension. In this embodiment, the type of the other object on which the topology structure is constructed should be consistent with the type of the target object, for example, if the target object is a vehicle, then the other object is also a vehicle.

And S503, using the candidate point position with the characteristics matched with the characteristics of the plurality of historical point positions in the plurality of candidate point positions as the point position at the next moment of the sequence to be predicted.

The matching rules may be different according to different application scenarios. For example, in one possible embodiment, for each candidate point location of the plurality of candidate point locations, a similarity between a feature of the candidate point location and features of the plurality of historical point locations may be calculated. And regarding the candidate point position with the highest similarity in the plurality of candidate point positions as the candidate point position matched with the characteristics of the plurality of historical point positions, and taking the candidate point position as the point position at the next moment of the sequence to be predicted. The similarity may be calculated by using an euclidean distance, a mahalanobis distance, a cosine similarity, and the like, which is not limited in this embodiment of the present invention.

The similarity of the features of one candidate point location and the plurality of historical point locations may reflect the overall degree of similarity between the candidate point location and the plurality of historical point locations, and the higher the degree of similarity is, the higher the possibility that the candidate point location is the point location at the next moment of the sequence to be predicted is, and the lower the degree of similarity is, the lower the possibility that the candidate point location is the point location at the next moment of the sequence to be predicted is. In another possible embodiment, the candidate point locations may also be clustered according to the features of the candidate point locations and the features of the plurality of historical point locations to obtain a plurality of candidate point location sets, a candidate point location set most likely to include a point location at a next time is selected from the plurality of candidate point location sets, for each candidate point location in the candidate point location set, each candidate point location is calculated by using a long-term and short-term network trained in advance as a score of the point location at the next time, and the candidate point location with the highest score is regarded as a candidate point location matched with the features of the plurality of historical point locations and is regarded as the point location at the next time of the sequence to be predicted.

The following describes the construction of the topological graph structure used in the trajectory prediction method provided in the embodiment of the present invention, and since the construction is basically similar to the topological graph structure used in the foregoing traffic prediction method, the relevant points can be referred to the foregoing description about the construction of the topological graph structure, and may be referred to fig. 6, which includes:

s601, a plurality of bayonet data of the target object and at least one other object are obtained.

Taking traffic flow predictions as an example, it can be appreciated that each vehicle is affected by the subjective factors of the driver, and the subjective concepts of different drivers may be different, so the characteristics of the travel path may not be identical to those of other vehicles, e.g., there are two roads between location a and location D, one road via location B, denoted as a-B-D, one road via location C, denoted as a-C-D, the driver of vehicle 1 may prefer to select roads a-B-D, and the driver of vehicle 2 may prefer to select roads a-C-D. Meanwhile, the driving tracks of the vehicles can be influenced by objective factors such as weather and roads, so that the characteristics of the driving tracks of different vehicles can have some commonality, for example, most drivers tend to select roads a-B-D on the assumption that the road surfaces of the roads a-B-D are wider and the routes are shorter than the roads a-C-D. The bayonet data of the target object can reflect subjective characteristics of a driver of the target object to a certain degree, and the bayonet data of other objects can reflect some commonalities generated by the influence of objective factors on the vehicle track to a certain degree. Thus, when predicting the trajectory of the target object, the bayonet data of the target object and the other objects can be simultaneously referred to.

And S602, aggregating point positions corresponding to two pieces of bayonet data belonging to the same time window and the same object in the plurality of bayonet data to obtain the historical tracks of the target object and at least one other object.

For example, if the time difference threshold is 10 minutes, the bayonet data 1 and the bayonet data 2 belong to the same time window, and if the time difference threshold is 5 minutes, the bayonet data 1 and the bayonet data 2 do not belong to the same time window. The time corresponding to the bayonet data refers to the time when the bayonet data is acquired.

Belonging to the same object means that two pieces of bayonet data are obtained by collecting the same object, and the case of representing the bayonet data in a key-id-value form is taken as an example, if the ids of the two pieces of bayonet data are the same, the two pieces of bayonet data can be regarded as belonging to the same object.

S603, constructing a topological graph structure based on point location pairs formed by adjacent point locations in the historical track of the target object, and obtaining an individual topological graph structure.

For how to obtain the topological graph based on the point pair construction, reference may be made to the foregoing description about the topological graph structure construction, and details are not described here again. As with the foregoing analysis, the individual topology may reflect, to some extent, the characteristics of the target object as distinguished from other objects.

S604, constructing a topological graph structure based on point position pairs formed by adjacent point positions in the historical track of at least one other object to obtain a group topological graph structure.

As with the foregoing analysis, the population topology may reflect to some extent the commonality of the target object and other objects. It is understood that fig. 6 is only a flow for constructing a topology structure in one possible embodiment, and in other possible embodiments, S604 may be executed before S603, or may be executed in parallel with S603 or executed alternately, which is not limited in this embodiment.

And S605, fusing the individual topological graph structure and the group topological graph structure.

The fusion mode can use the same type of edge fusion or different types of edge fusion. The same type of edge fusion may be to recalculate the edge weight of the edge in the topology structure obtained by fusion according to the following formula:

wherein P (B | A) is the edge weight of the edge between the node B and the node A in the topology graph structure obtained by fusion, P_{Group of people}(B | A) is the edge weight of the edge between node B and node A in the group topology, P_Individuals(B | A) is an edge weight of an edge between node B and node A in the individual topology structure. P_{Group of people}(I | A) is the edge weight of the edge between node I and node A in the group topology, P_Individuals(I | A) is the edge weight of the edge between node I and node A in the individual topological graph structure, node I is any node adjacent to node A, and neighbor represents that the summation range for I is all nodes adjacent to node A.

Some people may substitute a legal license plate with an illegal license plate in order to evade supervision, so as to disguise the vehicle, which is hereinafter referred to as a fake-licensed vehicle. In order to effectively detect a fake-licensed vehicle, an embodiment of the present invention provides a fake-licensed vehicle detection method, and referring to fig. 7, fig. 7 is a schematic flow diagram of the fake-licensed vehicle detection method provided by the embodiment of the present invention, and the method may include:

s701, acquiring a historical track of the vehicle to be detected.

The vehicle to be detected is not particularly limited to one or more vehicles, and may be different according to different application scenarios, which is not limited in this embodiment.

S702, constructing a topological graph structure based on historical track point positions in the historical tracks.

Regarding the construction of the topology structure, the detailed description will be made in the following embodiments, and the detailed description is omitted here.

S703, calculating the number of connected subgraphs in the topological graph structure.

The connected subgraph is a subset of nodes and edges in the topological graph structure, any two nodes in each connected subgraph are connected through the edges in the connected subgraph, and each node belongs to one connected subgraph only. For example, assuming that a topology structure includes 8 nodes in total, the number of nodes a to H is 8, in one possible embodiment, the number of edges between the nodes may be as shown in fig. 8a, it is known that the nodes a to H and all the edges together form a connected subgraph, that is, the number of connected subgraphs in this embodiment is 1, in another possible embodiment, the number of edges between the nodes may be as shown in fig. 8b, it is known that the nodes a to D and the edges AB, AC, BD, and CD form a connected subgraph, and the nodes E to H and the edges EF, FG, and GH form a connected subgraph, that is, the number of connected subgraphs in this embodiment is 2. The number of connected subgraphs can be calculated by using a community discovery algorithm in the related art.

S704, it is determined whether the number is 1.

S705, if the number is not 1, determining that the vehicle to be detected is a fake plate vehicle.

It can be understood that two adjacent nodes in the topological graph structure belong to the same connected subgraph, the topological graph structure is constructed based on the historical track of the vehicle to be detected, theoretically, the driving track of the vehicle should be continuous, therefore, the nodes corresponding to adjacent point positions in the driving track are adjacent theoretically in the topological graph structure, and therefore, all the nodes in the topological graph structure belong to the same connected subgraph, namely, the number of the connected subgraphs is 1. Due to the fact that other vehicles are caused by replacement of fake license plates in part of time, the obtained driving track of the fake-licensed vehicle is not continuous, namely a plurality of connected subgraphs may exist in the topological graph structure. Therefore, by adopting the embodiment, whether the vehicle to be detected is the fake-licensed vehicle can be detected by judging whether the number of the communicated subgraphs is 1.

A process of constructing a topological graph structure in the fake-licensed vehicle detection method according to the embodiment of the present invention is described below, and referring to fig. 9, fig. 9 is a schematic flowchart of a process of constructing a topological graph structure applied to fake-licensed vehicle detection according to the embodiment of the present invention, and the process may include:

s901, determining the times of each historical track point in the historical track aiming at each historical track point in the historical track.

For example, assuming that a total of 3 historical tracks, namely a-B-C-D, a-F-D-B and B-C-E-G, are included, the number of occurrences of the historical track point a is 2, and the number of occurrences of the historical track point B is 3. For the historical track, reference may be made to the manner of determining the sequence to be predicted in the foregoing S501, and it may be understood that the sequence to be predicted may be regarded as a track.

And S902, determining whether the frequency of the historical track point location is greater than a preset frequency threshold value.

The preset times threshold value can be set according to actual experience or requirements, and it can be understood that the point locations with a large number of occurrences can be regarded as the point locations where the vehicle to be detected frequently appears, and the point locations with a small number of occurrences can be regarded as the point locations where the vehicle to be detected rarely appears.

And S903, if the frequency is greater than a preset frequency threshold, taking the historical track point location as a hot historical track point location.

If the vehicle to be detected frequently appears at one point location, the credibility of the point location can be considered to be high, namely the higher probability of the vehicle to be detected actually appears at the point location, if the vehicle to be detected rarely appears at one point location, the credibility of the point location can be considered to be low, namely the vehicle to be detected has a certain probability and does not actually appear at the point location, and the influence that the vehicle to be detected possibly appears at the point location and is influenced by the interference signal is obtained.

And S904, constructing a topological graph structure based on all hot historical track point positions.

Reference may be made to the foregoing description on the topology construction, which is not described herein again.

Referring to fig. 10, fig. 10 is a schematic structural diagram of a flow predicting apparatus according to an embodiment of the present invention, which may include:

the feature obtaining module 1001 is configured to obtain features of a point to be predicted in multiple dimensions, where the features in the multiple dimensions at least include a topological graph structure feature and a historical flow feature, the topological graph structure feature is used to represent a connection relationship between the point to be predicted and other points in the topological graph structure, and the topological graph structure is constructed based on internet of things data collected at the point to be predicted and other points and is used to represent relevance of the point to be predicted and other points in a time dimension and/or a space dimension;

the flow prediction module 1002 is configured to input features of a point location to be predicted on multiple dimensions to a flow prediction model, so as to obtain a flow prediction result of the point location to be predicted, where the flow prediction model is trained in advance by sample data labeled with a true value, and the sample data is features of the point location to be predicted and multiple point locations among other point locations labeled with true values on the multiple dimensions.

In one possible embodiment, the topology is constructed by:

acquiring a plurality of pieces of internet of things data acquired at a point location to be predicted and other point locations, wherein each piece of internet of things data comprises a plurality of pieces of data in dimensionality, and at least comprises a piece of data in a time dimensionality and a piece of data in a space dimensionality;

constructing a plurality of nodes according to the fractional data of the Internet of things on the spatial dimension, wherein each node is used for representing a point location to be predicted and one point location of other point locations;

determining the correlation between every two nodes in the plurality of nodes, wherein the correlation is used for representing the data of the internet of things corresponding to one node in the two nodes, and the correlation degree between the data of the internet of things corresponding to the other node in the space dimension and the time dimension;

and according to the correlation among the nodes, constructing edges among the nodes to obtain the topological graph structure.

In one possible embodiment, determining a correlation between each two of the plurality of nodes comprises:

determining the interval of the data of the internet of things corresponding to one node in the two nodes and the data of the internet of things corresponding to the other node in the space dimension and/or the time dimension aiming at every two nodes in the plurality of nodes;

and if the interval is smaller than a preset interval threshold, determining that the correlation between the two nodes is related.

In one possible embodiment, the flow prediction model is trained in advance by:

and training to obtain a flow prediction model by utilizing a plurality of time sequence data of each node in the topological graph structure.

In a possible embodiment, the training to obtain the traffic prediction model by using a plurality of time series data of each node in the topological graph structure includes:

respectively training a plurality of basic models by utilizing the fractional data of a plurality of time sequence data of each node in the topological graph structure on different dimensions;

and fusing the trained basic models to obtain a flow prediction model.

Referring to fig. 11, fig. 11 is a schematic structural diagram of a trajectory prediction apparatus according to an embodiment of the present invention, which may include:

a prediction sequence module 1101, configured to determine a plurality of history point locations through which a target object passes, as a sequence to be predicted;

the feature expression module 1102 is configured to obtain features of each point location in the multiple point locations represented by the topological graph structure, where the features at least include topological graph structure features, the topological graph structure features are used to represent a connection relationship between the point location and other point locations in the topological graph structure, and the topological graph structure is constructed based on historical trajectories of the target object and at least one other object and is used to represent relevance of the multiple point locations in a time dimension and/or a space dimension;

the feature matching module 1103 is configured to use, as a next time point of the sequence to be predicted, a candidate point of the plurality of candidate points whose features are matched with the features of the plurality of historical points.

In a possible embodiment, the feature matching module 1103 is specifically configured to, for each candidate point location in the multiple candidate point locations, calculate similarity between features of the candidate point location and features of the multiple historical point locations;

and taking the candidate point position with the highest similarity in the candidate point positions as the next moment point position of the sequence to be predicted.

In one possible embodiment, the topology is constructed by:

acquiring a plurality of bayonet data of a target object and at least one other object;

aggregating point positions corresponding to two pieces of bayonet data of a plurality of bayonet data belonging to the same time window and the same object to obtain a historical track of a target object and at least one other object;

constructing a topological graph structure based on a point location pair consisting of adjacent point locations in the historical track of at least one other object to obtain a group topological graph structure;

Referring to fig. 12, fig. 12 is a schematic structural diagram of a fake-licensed vehicle detecting device according to an embodiment of the present invention, which may include:

a track obtaining module 1201, configured to obtain a historical track of a vehicle to be detected;

a topological graph construction module 1202, configured to construct a topological graph structure based on historical track point locations in the historical tracks;

a connected subgraph calculation module 1203, configured to calculate the number of connected subgraphs in the topology graph structure, where a connected subgraph is a subset of nodes and edges in the topology graph structure, and any two nodes in each connected subgraph are connected by an edge in the connected subgraph;

a judging module 1204, configured to determine whether the number is 1;

and the detecting module 1205 is used for determining that the vehicle to be detected is a fake plate vehicle if the number is not 1.

In one possible embodiment, the detection module 1205 is further configured to determine that the vehicle to be detected is not a fake-licensed vehicle if the number is 1.

In a possible embodiment, the topological graph constructing module 1202 is specifically configured to determine, for each historical trajectory point location in the historical trajectory, the number of times that the historical trajectory point location is in the historical trajectory; determining whether the frequency of the historical track point location is greater than a preset frequency threshold value; if the times are larger than a preset time threshold value, taking the historical track point location as a hot historical track point location; and constructing a topological graph structure based on all hot historical track point positions.

An embodiment of the present invention further provides an electronic device, as shown in fig. 13, including:

a memory 1301 for storing a computer program;

when the processor 1302 is configured to execute the program stored in the memory 1301, the electronic device may perform the following steps:

acquiring characteristics of a point to be predicted on multiple dimensions, wherein the characteristics on the multiple dimensions at least comprise topological graph structure characteristics and historical flow characteristics, the topological graph structure characteristics are used for representing the communication relation between the point to be predicted and other points in a topological graph structure, and the topological graph structure is constructed on the basis of Internet of things data acquired from the point to be predicted and other points and is used for representing the relevance of the point to be predicted and other points on a time dimension and/or a space dimension;

inputting the characteristics of the point location to be predicted on multiple dimensions into a flow prediction model to obtain a flow prediction result of the point location to be predicted, wherein the flow prediction model is trained by sample data marked with truth values in advance, and the sample data is the characteristics of the point location to be predicted and multiple point locations of other point locations marked with truth values on the multiple dimensions.

When the electronic device is used for trajectory prediction, the following steps can be realized:

the method comprises the steps of obtaining characteristics of a plurality of historical point positions and a plurality of candidate point positions, wherein the characteristics at least comprise topological graph structure characteristics, the topological graph structure characteristics are used for representing the connection relation between the point position and other point positions in a topological graph structure, and the topological graph structure is constructed on the basis of historical tracks of a target object and at least one other object and is used for representing the relevance of the plurality of historical point positions and the plurality of candidate point positions in a time dimension and/or a space dimension;

and taking the candidate point location with the characteristics matched with the characteristics of the plurality of historical point locations in the plurality of candidate point locations as the point location at the next moment of the sequence to be predicted.

When the electronic equipment is used for detecting the fake-licensed vehicle, the following steps can be realized:

acquiring a plurality of history point positions passed by a target object and at least one other object;

aggregating every two historical point positions belonging to the same time window and the same object in the plurality of historical point positions to obtain the historical tracks of the target object and at least one other object;

constructing a topological graph structure based on a historical point location pair formed by adjacent historical point locations in the historical track of the target object to obtain an individual topological graph structure;

constructing a topological graph structure based on a historical point location pair formed by adjacent historical point locations in the historical track of at least one other object to obtain a group topological graph structure;

The aforementioned electronic device may include a Random Access Memory (RAM) and a Non-Volatile Memory (NVM), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the processor.

The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but also Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components.

In yet another embodiment of the present invention, a computer-readable storage medium is further provided, which has instructions stored therein, and when the instructions are executed on a computer, the instructions cause the computer to execute any one of the above-mentioned flow prediction methods.

In yet another embodiment, a computer program product containing instructions is provided, which when run on a computer, causes the computer to perform any of the above described methods of flow prediction.

In yet another embodiment of the present invention, a computer-readable storage medium is further provided, which has instructions stored therein, and when the instructions are executed on a computer, the instructions cause the computer to execute any of the trajectory prediction methods of the above embodiments.

In yet another embodiment, a computer program product containing instructions is provided, which when run on a computer, causes the computer to perform any of the trajectory prediction methods of the above embodiments.

In yet another embodiment of the present invention, a computer-readable storage medium is provided, having instructions stored thereon, which when run on a computer, cause the computer to perform any one of the methods of fake-licensed vehicle detection described above.

In yet another embodiment of the present invention, there is also provided a computer program product containing instructions which, when run on a computer, cause the computer to perform any of the methods of the above embodiments of fake-licensed vehicle detection.

In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.

It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.

All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the embodiments of the apparatus, the electronic device, the computer-readable storage medium, and the computer program product, since they are substantially similar to the method embodiments, the description is relatively simple, and for the relevant points, reference may be made to the partial description of the method embodiments.

The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method of traffic prediction, the method comprising:

2. The method of claim 1, wherein the topological graph structure is constructed by:

3. The method of claim 2, wherein determining the correlation between each two of the plurality of nodes comprises:

4. The method of claim 2, wherein the flow prediction model is previously trained by:

5. The method of claim 4, wherein training a traffic prediction model using the plurality of time series data of each node in the topological graph structure comprises:

and fusing the trained basic models to obtain a flow prediction model.

6. A trajectory prediction method, characterized in that the method comprises:

7. The method according to claim 6, wherein the regarding the candidate point location of the plurality of candidate point locations, where the feature is matched with the feature of the plurality of historical point locations, as the next point location of the sequence to be predicted, comprises:

8. The method of claim 6, wherein the topological graph structure is constructed by:

9. A flow prediction apparatus, characterized in that the apparatus comprises:

10. A trajectory prediction device, characterized in that the device comprises: